Hi, I want to configure spicedb dispatching without kubenretes, but documentation is not very specific, am I right that I need something like nginx with all slicedb dispatcher upstreams? Maybe there are and article or example about configuring dispatcher with k8s?

Hi, dispatching with Kubernetes API is not documented and experimental, we haven't seen uses of it in production. The only alternative to the discovering peers via the Kubernetes API is using XDS: https://www.envoyproxy.io/docs/envoy/latest/api-docs/xds_protocol

Ok, thank you, looks complicated, maybe we do not need dispatching because our schema should be simple. Am I right that for configuring simple high availability I can start two instances of spicedb on one datastore, and some proxy before it?

dispatch is an integral part to SpiceDB's / Zanzibar model. It largely depends on your schema and data loaded into the cluster, we've seen some folks running it without dispatch in production and work well for them. But please note it won't be leveraging the foundations for horizontal scalability. And yes, you can run a highly available cluster with N replicas and 1 datastore, behind a load balancer.